Reinforcement learning improves 3D gene mapping from fewer samples
STRATA: Reinforcement-driven Active Sampling with Tri-paradigm Adaptation for Efficient 3D Spatial Transcriptomics
Computer Vision and Pattern Recognition
Summary
Mapping where genes are active inside 3D tissue is expensive and slow. The authors created STRATA, an AI method that selects fewer but smarter spots to measure based on tissue images and single cell data, saving resources. Their approach combines different ways to predict missing gene data and adapts based on what measurements it has. Tests show STRATA can build detailed gene activity maps more accurately with less data.
What this means in practice
- •For biomedical imaging teams: Reduce total measurements needed to create detailed 3D gene activity maps in tissues using AI-driven spot selection.
- •For bioinformatics software developers: Develop tools that integrate morphology and sampling strategies to improve gene expression map accuracy from sparse data.
Authors
Junchao Zhu, Ruining Deng, Chongyu Qu, Zhengyi Lu, Yanfan Zhu, Yuechen Yang, Yu Wang, Shilin Zhao, Haichun Yang, Yuankai Huo
Abstract
Spatial transcriptomics (ST) enables joint characterization of tissue morphology and gene expression. However, 3D volumetric ST acquisition remains expensive. AI-driven methods offer a low-cost alternative but remain limited in accuracy, while recent 2D section-wise ST acquisition still rely on dense grid sampling, resulting in redundant measurements and undersampling of biologically informative regions. To bridge the gap, we propose STRATA, a reinforcement-driven active sampling framework with tri-paradigm adaptation for reconstructing dense 3D transcriptomic maps from sparse voxel-wise ST acquisition. STRATA constructs a 3D graph integrating H&E morphology with single-cell priors and employs a reinforcement learning agent to estimate morphological sparsity and transcriptomic informativeness, enabling budget-aware selection of informative spots. Given these sparse measurements, an adaptive tri-paradigm predictor combines retrieval, regression, and cross-layer imputation through lightweight fusion, dynamically weighting their contributions. Experiments across datasets demonstrate improved sampling efficiency and accuracy, particularly under low-budget settings. Code is available at https://github.com/Junchao-Zhu/STRATA.